Skip to main content

Building a Tag-Driven Metadata System for Political Domain Classification

· 2 min read
samundrak
JavaScript Dev

Overview​

As my civic-tech platform evolved, the complexity of classification increased: I needed to track ministry types, cabinet roles, party types (left, regional, royalist), election kinds, and more. Rather than hardcoding categories, I built a tag-driven metadata system.

This post explains the design and use cases for tag-based classification across multiple resources.

Use Cases for Tags​

Tags allowed flexible classification of:

  • Ministries (e.g., “Infrastructure”, “Defense”, “Education”)
  • Parties (e.g., “Leftist”, “Royalist”, “Regional”, “New”)
  • Governments (e.g., “Caretaker”, “Interim”)
  • Scandals (e.g., “Corruption”, “Misuse of Funds”)
  • Leaders (e.g., “Military Background”, “Convicted”)

Schema Design​

Tags are defined in a central table:

model Tag {
  id    Int    @id @default(autoincrement())
  name String
  nameLocal String?
  type String // e.g. 'MINISTRY_TYPE', 'PARTY_TYPE', 'LEADER_FLAG'
}

model TagLink {
  id           Int     @id @default(autoincrement())
  tagId        Int
  resourceType String  // 'PARTY', 'LEADER', 'GOVERNMENT', etc.
  resourceId   Int
}

This design supports many-to-many relations across arbitrary resource types.

Frontend Usage​

Tags are used in listings and detail pages. Example use:

{tags.map(tag => (
  <Badge key={tag.id}>{i18n.language === 'np' ? tag.nameLocal : tag.name}</Badge>
))}

They’re filterable in search and used to group similar entries across the app.

Dynamic Filtering​

Tags also drive dynamic filters in dashboards. For example:

filters: [
  { label: 'Leftist Parties', value: 'LEFTIST' },
  { label: 'Royalist', value: 'ROYALIST' }
]

Tag Suggestions + Admin Tools​

Admins can:

  • Suggest tags while editing a resource
  • View and manage global tag lists
  • Localize tag names from admin panel

Benefits​

This system allowed:

  • Dynamic taxonomy management
  • Reusable logic across many models
  • Easy localization
  • Classification at scale

Summary​

Tags replaced hardcoded booleans and enums with flexible metadata. They power filtering, grouping, labeling, and classification across the civic data ecosystem.

In the next article, I’ll break down how I used full-text search for discovering parties, leaders, elections, and more in both English and Nepali.